Re: [HTCondor-devel] DEFAULT_UPDATE_INTERVAL?

[Todd Tannenbaum <tannenba@xxxxxxxxxxx>]

Re the below, perhaps just a metaknob (see https://is.gd/IgD7MV ), so 
the admin could just say something like
  use policy : fast_polling
or
  use policy : scalable_polling

?  We could come up with better names.

By using a metaknob, we
   a) don't have to change any code, just add it into param_info.in 
(see https://is.gd/m9xYSq )
   b) admins can query what the setting changes via condor_config_val 
use policy : fast_polling
just my $.02,
Todd


On 6/7/2016 2:26 PM, Ben Cotton wrote:
> Hi friends,
>
> In some of my work with customers, they've expressed a desire to
> shorten the update interval all around. The two main motivations are:
>    1. cloud environments, where shortening CLASSAD_LIFETIME means nodes
> that go poof unexpectedly are dropped from the collector sooner
>    2. HA failover using an external tool (e.g. Pacemaker) has shorter
> missing host times (we always suggest using HTCondor's HA mechanism,
> but...)
>
> Before I went off and filed an enhancement request in Trac, I was
> curious if this had been considered and rejected before. I couldn't
> find a relevant ticket, but that doesn't mean it's not there (or
> perhaps it never made it that far).
>
> What I have in mind is a configuration macro called
> DEFAULT_UPDATE_INTERVAL which will serve as the value of the following
> when they are not explicitly set in a config file:
>      UPDATE_INTERVAL
>      NEGOTIATOR_UPDATE_INTERVAL
>      SCHEDD_INTERVAL
>      MASTER_UPDATE_INTERVAL
>      CKPT_SERVER_INTERVAL
>      DEFRAG_UPDATE_INTERVAL
>      HAD_UPDATE_INTERVAL
>
> While there are certainly cases where setting these individually make
> sense, it would be nice to be able to change them with a single edit
> since they all have the same default.
>
> An alternate idea would be to deprecate all of those in favor of
> simply using UPDATE_INTERVAL (or CLASSAD_UPDATE_INTERVAL to make it
> more clear that it's a new setting) and then individual daemons can be
> changed with, e.g., STARTD.UPDATE_INTERVAL. That could be a pretty
> significant change for people, so it's probably not worth the effort.
>
>
> Thanks,
> BC

--
Todd Tannenbaum <tannenba@xxxxxxxxxxx> University of Wisconsin-Madison
Center for High Throughput Computing   Department of Computer Sciences
HTCondor Technical Lead                1210 W. Dayton St. Rm #4257
Phone: (608) 263-7132                  Madison, WI 53706-1685